What happened
On 7 October 2026 OpenAI published its deployment-safety system card for GPT-6 Sol and GPT-6 Luna as the models rolled out to all ChatGPT users globally, updating prior September versions. The card discloses that 'GPT-6 in ChatGPT incorporates Astra's safety advances' and, versus the GPT-5.6 models, 'showed stronger resistance to jailbreaks, including attacks that adapt across multiple turns, as well as reductions in dishonesty, deception, and circumvention of guardrails'; areas with regressions were manually reviewed and red-teamed and 'found that the disallowed responses were generally low severity.' Under the Preparedness Framework, the October release is treated as 'High capability in both Cybersecurity and Biological and Chemical domains,' while neither model reaches High in AI Self-Improvement, so the release carries the same safeguards detailed in the GPT-5.6 system card.
Why it matters
The card is the authoritative statement of a frontier lab's current safety-and-capability posture on its flagship models — including which cyber and biological misuse thresholds the company treats as high-risk — useful benchmark context for security, compliance and procurement teams evaluating frontier-model risk.
Action needed
Use the prepared assessments (Cyber and BioChem High capability classifications) to calibrate your enterprise guardrail, red-team and acceptable-use assumptions for GPT-6 against OpenAI's own stated threat model.